hes@ecsvax.UUCP (03/25/84)
<> I've been interested in the relationship between mobility of DNA fragments on agarose gels and their lengths. The original method (and still used by some) was to plot mobilities and lengths for some known fragments on semi-log paper, and hope that a straight line resulted. One could then fit a straight line by linear regression, and use it to predict the lengths from mobilities of unknown fragments. The main problem is that a curved line usually results, and so the custom was to use linear regression for the straight region(s), and eyeball the curved region. We were able to put the reciprocal model of Southern in a statistical least squares framework, and end up with a procedure which both fits the data much better than the semi-log procedure, and also gives some statistical type insights into error (lack-of-fit.) This is an easy computation for a computer, but not inviting for hand computation. Our work is published as "Improved Estimation of DNA Fragment Lengths from Agarose Gels" in Analytical Biochemistry 115:113-122(1981) This paper includes a FORTRAN program for the analysis. That program is batch oriented, and I have a BASIC interactive version which is the one we use. If there are requests, I'd be glad to mail it to you or put it in net.sources. --henry schaffer genetics ncsu